Poverty Level Analysis
R Markdown
This is an R Markdown document. Markdown is a simple formatting syntax for authoring HTML, PDF, and MS Word documents. For more details on using R Markdown see http://rmarkdown.rstudio.com.
When you click the Knit button a document will be generated that includes both content as well as the output of any embedded R code chunks within the document. You can embed an R code chunk like this:
Loading Libraries
## Warning: package 'tidyverse' was built under R version 4.0.2## ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.1 ──## ✓ ggplot2 3.3.5 ✓ purrr 0.3.4
## ✓ tibble 3.1.4 ✓ dplyr 1.0.7
## ✓ tidyr 1.1.3 ✓ stringr 1.4.0
## ✓ readr 2.0.1 ✓ forcats 0.5.1## Warning: package 'ggplot2' was built under R version 4.0.2## Warning: package 'tidyr' was built under R version 4.0.2## Warning: package 'readr' was built under R version 4.0.2## Warning: package 'purrr' was built under R version 4.0.2## Warning: package 'dplyr' was built under R version 4.0.2## Warning: package 'stringr' was built under R version 4.0.2## Warning: package 'forcats' was built under R version 4.0.2## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()## Warning: package 'xgboost' was built under R version 4.0.2##
## Attaching package: 'xgboost'## The following object is masked from 'package:dplyr':
##
## slice## Warning: package 'caret' was built under R version 4.0.2## Loading required package: lattice##
## Attaching package: 'caret'## The following object is masked from 'package:purrr':
##
## lift## Warning: package 'VIM' was built under R version 4.0.2## Loading required package: colorspace## Warning: package 'colorspace' was built under R version 4.0.2## Loading required package: grid## VIM is ready to use.## Suggestions and bug-reports can be submitted at: https://github.com/statistikat/VIM/issues##
## Attaching package: 'VIM'## The following object is masked from 'package:datasets':
##
## sleep## corrplot 0.90 loaded## Warning: package 'ggthemes' was built under R version 4.0.2## Warning: package 'plotly' was built under R version 4.0.2##
## Attaching package: 'plotly'## The following object is masked from 'package:xgboost':
##
## slice## The following object is masked from 'package:ggplot2':
##
## last_plot## The following object is masked from 'package:stats':
##
## filter## The following object is masked from 'package:graphics':
##
## layout## Warning: package 'factoextra' was built under R version 4.0.2## Welcome! Want to learn more? See two factoextra-related books at https://goo.gl/ve3WBa## Warning: package 'Rtsne' was built under R version 4.0.2## Warning: package 'breakDown' was built under R version 4.0.2##
## Attaching package: 'breakDown'## The following object is masked from 'package:VIM':
##
## wine## Warning: package 'gridExtra' was built under R version 4.0.2##
## Attaching package: 'gridExtra'## The following object is masked from 'package:dplyr':
##
## combine## Warning: package 'lightgbm' was built under R version 4.0.2## Loading required package: R6## Warning: package 'R6' was built under R version 4.0.2##
## Attaching package: 'lightgbm'## The following object is masked from 'package:plotly':
##
## slice## The following objects are masked from 'package:xgboost':
##
## getinfo, setinfo, slice## The following object is masked from 'package:dplyr':
##
## slice## Warning: package 'glmnet' was built under R version 4.0.2## Loading required package: Matrix##
## Attaching package: 'Matrix'## The following objects are masked from 'package:tidyr':
##
## expand, pack, unpack## Loaded glmnet 4.1-2Importing Data
train = read_csv("./Data/train.csv")## Rows: 9557 Columns: 143## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (5): Id, idhogar, dependency, edjefe, edjefa
## dbl (138): v2a1, hacdor, rooms, hacapo, v14a, refrig, v18q, v18q1, r4h1, r4h...##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.test = read_csv("./Data/test.csv")## Rows: 23856 Columns: 142## ── Column specification ────────────────────────────────────────────────────────
## Delimiter: ","
## chr (5): Id, idhogar, dependency, edjefe, edjefa
## dbl (137): v2a1, hacdor, rooms, hacapo, v14a, refrig, v18q, v18q1, r4h1, r4h...##
## ℹ Use `spec()` to retrieve the full column specification for this data.
## ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.cat("Dimensions for training data : ", dim(train))## Dimensions for training data : 9557 143cat("Dimensions for testing data: ", dim(test))## Dimensions for testing data: 23856 142head(train)## # A tibble: 6 × 143
## Id v2a1 hacdor rooms hacapo v14a refrig v18q v18q1 r4h1 r4h2 r4h3
## <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 ID_2796… 190000 0 3 0 1 1 0 NA 0 1 1
## 2 ID_f29e… 135000 0 4 0 1 1 1 1 0 1 1
## 3 ID_68de… NA 0 8 0 1 1 0 NA 0 0 0
## 4 ID_d671… 180000 0 5 0 1 1 1 1 0 2 2
## 5 ID_d56d… 180000 0 5 0 1 1 1 1 0 2 2
## 6 ID_ec05… 180000 0 5 0 1 1 1 1 0 2 2
## # … with 131 more variables: r4m1 <dbl>, r4m2 <dbl>, r4m3 <dbl>, r4t1 <dbl>,
## # r4t2 <dbl>, r4t3 <dbl>, tamhog <dbl>, tamviv <dbl>, escolari <dbl>,
## # rez_esc <dbl>, hhsize <dbl>, paredblolad <dbl>, paredzocalo <dbl>,
## # paredpreb <dbl>, pareddes <dbl>, paredmad <dbl>, paredzinc <dbl>,
## # paredfibras <dbl>, paredother <dbl>, pisomoscer <dbl>, pisocemento <dbl>,
## # pisoother <dbl>, pisonatur <dbl>, pisonotiene <dbl>, pisomadera <dbl>,
## # techozinc <dbl>, techoentrepiso <dbl>, techocane <dbl>, techootro <dbl>, …Encoding categorical features (Already one-hot encoded, reversing present encoding, adding new integer encoding)
One hot encoding is one method of converting data to prepare it for an algorithm and get a better prediction. With one-hot, we convert each categorical value into a new categorical column and assign a binary value of 1 or 0 to those columns. Each integer value is represented as a binary vector.
Cycle through and reverse the OHE process for these
## pared_int piso_int techo_int abasta_int sanitario_int energcocinar_int
## 1 1 3 2 1 2 3
## 2 4 2 4 1 2 2
## 3 4 3 4 1 2 2
## 4 1 3 4 1 2 2
## 5 1 3 4 1 2 2
## 6 1 3 4 1 2 2
## elimbasu_int epared_int etecho_int eviv_int estadocivil_int parentesco_int
## 1 1 2 1 1 4 1
## 2 1 2 2 2 4 1
## 3 1 2 3 3 6 1
## 4 1 3 3 3 7 6
## 5 1 3 3 3 2 5
## 6 1 3 3 3 2 1
## instlevel_int tipovivi_int lugar_int area_int
## 1 4 3 1 1
## 2 8 3 1 1
## 3 5 1 1 1
## 4 4 3 1 1
## 5 5 3 1 1
## 6 5 3 1 1## Id v2a1 hacdor rooms hacapo v14a refrig v18q v18q1 r4h1 r4h2 r4h3
## 1 ID_279628684 190000 0 3 0 1 1 0 NA 0 1 1
## 2 ID_f29eb3ddd 135000 0 4 0 1 1 1 1 0 1 1
## 3 ID_68de51c94 NA 0 8 0 1 1 0 NA 0 0 0
## 4 ID_d671db89c 180000 0 5 0 1 1 1 1 0 2 2
## 5 ID_d56d6f5f5 180000 0 5 0 1 1 1 1 0 2 2
## 6 ID_ec05b1a7b 180000 0 5 0 1 1 1 1 0 2 2
## r4m1 r4m2 r4m3 r4t1 r4t2 r4t3 tamhog tamviv escolari rez_esc hhsize
## 1 0 0 0 0 1 1 1 1 10 NA 1
## 2 0 0 0 0 1 1 1 1 12 NA 1
## 3 0 1 1 0 1 1 1 1 11 NA 1
## 4 1 1 2 1 3 4 4 4 9 1 4
## 5 1 1 2 1 3 4 4 4 11 NA 4
## 6 1 1 2 1 3 4 4 4 11 NA 4
## paredblolad paredzocalo paredpreb pareddes paredmad paredzinc paredfibras
## 1 1 0 0 0 0 0 0
## 2 0 0 0 0 1 0 0
## 3 0 0 0 0 1 0 0
## 4 1 0 0 0 0 0 0
## 5 1 0 0 0 0 0 0
## 6 1 0 0 0 0 0 0
## paredother pisomoscer pisocemento pisoother pisonatur pisonotiene pisomadera
## 1 0 1 0 0 0 0 0
## 2 0 0 0 0 0 0 1
## 3 0 1 0 0 0 0 0
## 4 0 1 0 0 0 0 0
## 5 0 1 0 0 0 0 0
## 6 0 1 0 0 0 0 0
## techozinc techoentrepiso techocane techootro cielorazo abastaguadentro
## 1 0 1 0 0 1 1
## 2 1 0 0 0 1 1
## 3 1 0 0 0 1 1
## 4 1 0 0 0 1 1
## 5 1 0 0 0 1 1
## 6 1 0 0 0 1 1
## abastaguafuera abastaguano public planpri noelec coopele sanitario1
## 1 0 0 1 0 0 0 0
## 2 0 0 1 0 0 0 0
## 3 0 0 1 0 0 0 0
## 4 0 0 1 0 0 0 0
## 5 0 0 1 0 0 0 0
## 6 0 0 1 0 0 0 0
## sanitario2 sanitario3 sanitario5 sanitario6 energcocinar1 energcocinar2
## 1 1 0 0 0 0 0
## 2 1 0 0 0 0 1
## 3 1 0 0 0 0 1
## 4 1 0 0 0 0 1
## 5 1 0 0 0 0 1
## 6 1 0 0 0 0 1
## energcocinar3 energcocinar4 elimbasu1 elimbasu2 elimbasu3 elimbasu4 elimbasu5
## 1 1 0 1 0 0 0 0
## 2 0 0 1 0 0 0 0
## 3 0 0 1 0 0 0 0
## 4 0 0 1 0 0 0 0
## 5 0 0 1 0 0 0 0
## 6 0 0 1 0 0 0 0
## elimbasu6 epared1 epared2 epared3 etecho1 etecho2 etecho3 eviv1 eviv2 eviv3
## 1 0 0 1 0 1 0 0 1 0 0
## 2 0 0 1 0 0 1 0 0 1 0
## 3 0 0 1 0 0 0 1 0 0 1
## 4 0 0 0 1 0 0 1 0 0 1
## 5 0 0 0 1 0 0 1 0 0 1
## 6 0 0 0 1 0 0 1 0 0 1
## dis male female estadocivil1 estadocivil2 estadocivil3 estadocivil4
## 1 0 1 0 0 0 0 1
## 2 0 1 0 0 0 0 1
## 3 1 0 1 0 0 0 0
## 4 0 1 0 0 0 0 0
## 5 0 0 1 0 1 0 0
## 6 0 1 0 0 1 0 0
## estadocivil5 estadocivil6 estadocivil7 parentesco1 parentesco2 parentesco3
## 1 0 0 0 1 0 0
## 2 0 0 0 1 0 0
## 3 0 1 0 1 0 0
## 4 0 0 1 0 0 1
## 5 0 0 0 0 1 0
## 6 0 0 0 1 0 0
## parentesco4 parentesco5 parentesco6 parentesco7 parentesco8 parentesco9
## 1 0 0 0 0 0 0
## 2 0 0 0 0 0 0
## 3 0 0 0 0 0 0
## 4 0 0 0 0 0 0
## 5 0 0 0 0 0 0
## 6 0 0 0 0 0 0
## parentesco10 parentesco11 parentesco12 idhogar hogar_nin hogar_adul
## 1 0 0 0 21eb7fcc1 0 1
## 2 0 0 0 0e5d7a658 0 1
## 3 0 0 0 2c7317ea8 0 1
## 4 0 0 0 2b58d945f 2 2
## 5 0 0 0 2b58d945f 2 2
## 6 0 0 0 2b58d945f 2 2
## hogar_mayor hogar_total dependency edjefe edjefa meaneduc instlevel1
## 1 0 1 no 10 no 10 0
## 2 1 1 8 12 no 12 0
## 3 1 1 8 no 11 11 0
## 4 0 4 yes 11 no 11 0
## 5 0 4 yes 11 no 11 0
## 6 0 4 yes 11 no 11 0
## instlevel2 instlevel3 instlevel4 instlevel5 instlevel6 instlevel7 instlevel8
## 1 0 0 1 0 0 0 0
## 2 0 0 0 0 0 0 1
## 3 0 0 0 1 0 0 0
## 4 0 0 1 0 0 0 0
## 5 0 0 0 1 0 0 0
## 6 0 0 0 1 0 0 0
## instlevel9 bedrooms overcrowding tipovivi1 tipovivi2 tipovivi3 tipovivi4
## 1 0 1 1.000000 0 0 1 0
## 2 0 1 1.000000 0 0 1 0
## 3 0 2 0.500000 1 0 0 0
## 4 0 3 1.333333 0 0 1 0
## 5 0 3 1.333333 0 0 1 0
## 6 0 3 1.333333 0 0 1 0
## tipovivi5 computer television mobilephone qmobilephone lugar1 lugar2 lugar3
## 1 0 0 0 1 1 1 0 0
## 2 0 0 0 1 1 1 0 0
## 3 0 0 0 0 0 1 0 0
## 4 0 0 0 1 3 1 0 0
## 5 0 0 0 1 3 1 0 0
## 6 0 0 0 1 3 1 0 0
## lugar4 lugar5 lugar6 area1 area2 age SQBescolari SQBage SQBhogar_total
## 1 0 0 0 1 0 43 100 1849 1
## 2 0 0 0 1 0 67 144 4489 1
## 3 0 0 0 1 0 92 121 8464 1
## 4 0 0 0 1 0 17 81 289 16
## 5 0 0 0 1 0 37 121 1369 16
## 6 0 0 0 1 0 38 121 1444 16
## SQBedjefe SQBhogar_nin SQBovercrowding SQBdependency SQBmeaned agesq
## 1 100 0 1.000000 0 100 1849
## 2 144 0 1.000000 64 144 4489
## 3 0 0 0.250000 64 121 8464
## 4 121 4 1.777778 1 121 289
## 5 121 4 1.777778 1 121 1369
## 6 121 4 1.777778 1 121 1444
## pared_int piso_int techo_int abasta_int sanitario_int energcocinar_int
## 1 1 3 2 1 2 3
## 2 4 2 4 1 2 2
## 3 4 3 4 1 2 2
## 4 1 3 4 1 2 2
## 5 1 3 4 1 2 2
## 6 1 3 4 1 2 2
## elimbasu_int epared_int etecho_int eviv_int estadocivil_int parentesco_int
## 1 1 2 1 1 4 1
## 2 1 2 2 2 4 1
## 3 1 2 3 3 6 1
## 4 1 3 3 3 7 6
## 5 1 3 3 3 2 5
## 6 1 3 3 3 2 1
## instlevel_int tipovivi_int lugar_int area_int
## 1 4 3 1 1
## 2 8 3 1 1
## 3 5 1 1 1
## 4 4 3 1 1
## 5 5 3 1 1
## 6 5 3 1 1Introducing Ridge Encoding
# Grab OHE features
new_features_integer = data.frame(apply(new_features_integer,2,factor))
ridge_features = model.matrix(~.-1,new_features_integer[1:nrow(train),])
ridge_response = ifelse(train$Target == 4, 0, 1)
# Run ridge
cv.fit = cv.glmnet(
x = as.matrix(ridge_features),
y = as.factor(ridge_response),
nfolds = 5,
alpha = 0,
family = "binomial"
)
plot(cv.fit)
### Final preprocessing
# Rather than taking the predictions themselves, we'll take the coeffecients as encoding
tmp_coeffs = coef(cv.fit, s = "lambda.min")
coefs = data.frame(name = tmp_coeffs@Dimnames[[1]][tmp_coeffs@i + 1], coefficient = tmp_coeffs@x)
# Lookup
# Adjust original feature level names
for (i in 1:ncol(new_features_integer)){
new_features_integer[,i] = paste0(names(new_features_integer)[i], new_features_integer[,i])
}
new_features_ridge <- new_features_integer
new_features_ridge[] <- coefs$coefficient[match(unlist(new_features_integer), coefs$name)]
# Zeros can replace the NA since these are just base level
new_features_ridge[is.na(new_features_ridge)] = 0
# Rename for now ...
names(new_features_ridge) = gsub("int","ridge",names(new_features_ridge))
# Attach to data
full = data.frame(cbind(full,new_features_ridge), stringsAsFactors = FALSE)Visualization
# Filter the data for those features that actually have some missing data
full_missing = full[, sapply(full, anyNA), drop = FALSE]
cat("Missing data found in ", ncol(full_missing)/ncol(full)*100, "% of features")## Missing data found in 2.873563 % of features# Plot missing information for these features
aggr(full_missing, prop = T, numbers = T, cex.axis = 0.8)## Warning in plot.aggr(res, ...): not enough horizontal space to display
## frequencies
### Correlation
# Select numeric features (removing the categorical features for now and those with high % NA)
full_numerical = full %>%
select_if(is.numeric) %>%
select(-one_of(ohe_names)) %>%
select(
-v2a1,
-v18q1,
-rez_esc
)
# Since the number of NA is so low for the remaining features I will just replace them with their column mean
replace_na = function(x) replace(x, is.na(x), mean(x, na.rm = TRUE))
full_numerical = apply(full_numerical, 2, replace_na)
# Create the matrix
full_numerical_cor = cor(full_numerical, use = "complete.obs")
# Plot the matrix with some clustering
corrplot(full_numerical_cor, order = "hclust", tl.cex = 0.5)
# We have already created a numerical matrix with the categorical information excluded, so we will use this
full_pca = prcomp(full_numerical, center = TRUE, scale. = TRUE)
full = full %>%
cbind(full_pca$x[,1:3])
# scree plot
fviz_eig(full_pca)
# plot first two components
fviz_pca_var(full_pca,
col.var = "contrib", # Color by contributions to the PC
gradient.cols = c("#00AFBB", "#E7B800", "#FC4E07"),
select.var = list(contrib = 30), # top 30 contributing
repel = TRUE # Avoid text overlapping
)## Warning: ggrepel: 15 unlabeled data points (too many overlaps). Consider
## increasing max.overlaps
# Grab training set for visualisation
train_pca = full[1:nrow(train),] %>%
select(PC1,PC2,PC3) %>%
cbind(train$Target) %>%
rename(Target = `train$Target`) %>%
mutate(
Target = as.factor(Target),
Target_Binary = ifelse(Target == 4, "Non-vunerable", "Vunerable"))
# Plot
plot_ly(train_pca, x = ~PC1, y = ~PC2, z = ~PC3, color = ~Target) %>%
add_markers(marker = list(
size = 3,
opacity = 0.5)) %>%
layout(scene = list(xaxis = list(title = 'PC1'),
yaxis = list(title = 'PC2'),
zaxis = list(title = 'PC3'),
title = 'PCA plot (original target)'))plot_ly(train_pca, x = ~PC1, y = ~PC2, z = ~PC3, color = ~Target_Binary) %>%
add_markers(marker = list(
size = 3,
opacity = 0.5)) %>%
layout(scene = list(xaxis = list(title = 'PC1'),
yaxis = list(title = 'PC2'),
zaxis = list(title = 'PC3'),
title = 'PCA plot (binary target)'))## Warning in RColorBrewer::brewer.pal(N, "Set2"): minimal value for n is 3, returning requested palette with 3 different levels
## Warning in RColorBrewer::brewer.pal(N, "Set2"): minimal value for n is 3, returning requested palette with 3 different levels# Run tsne on sample of numerical features we used for PCA (training set only)
train_tsne = full_numerical[1:nrow(train),]
# Appears we have duplicates in our data - let's look
train_duplicates = which(duplicated(train_tsne))
# 23 instances: I suppose this could happen since we may have two adults in the same house with all numerical features being identical
# Remove dupes and scale
train_tsne_deduped = train_tsne[-train_duplicates,]
train_tsne_labels = train[-train_duplicates, "Target"]
train_tsne_labels = as.factor(train_tsne_labels$Target)
# Let's also include a binary version of our target to make visualisation easier
train_tsne_labels_binary = train[-train_duplicates, "Target"] %>%
mutate(
Target_Binary = ifelse(Target == 4, "Non-vunerable", "Vunerable")
)
train_tsne_labels_binary = as.factor(train_tsne_labels_binary$Target_Binary)
tsne = Rtsne(
train_tsne_deduped,
dims = 3,
perplexity = 50,
max_iter = 500,
verbose = TRUE,
theta = 0.35,
pca_scale = TRUE)## Performing PCA
## Read the 9544 x 50 data matrix successfully!
## Using no_dims = 3, perplexity = 50.000000, and theta = 0.350000
## Computing input similarities...
## Building tree...
## Done in 20.41 seconds (sparsity = 0.024120)!
## Learning embedding...
## Iteration 50: error is 90.364804 (50 iterations in 11.41 seconds)
## Iteration 100: error is 90.346986 (50 iterations in 16.57 seconds)
## Iteration 150: error is 89.100517 (50 iterations in 17.90 seconds)
## Iteration 200: error is 88.892110 (50 iterations in 12.29 seconds)
## Iteration 250: error is 88.876462 (50 iterations in 11.34 seconds)
## Iteration 300: error is 2.950109 (50 iterations in 7.90 seconds)
## Iteration 350: error is 2.553811 (50 iterations in 8.32 seconds)
## Iteration 400: error is 2.344003 (50 iterations in 8.99 seconds)
## Iteration 450: error is 2.207216 (50 iterations in 8.06 seconds)
## Iteration 500: error is 2.109347 (50 iterations in 8.51 seconds)
## Fitting performed in 111.29 seconds.# Plot with binary labels
tsne_plot_binary <- data.frame(x = tsne$Y[,1], y = tsne$Y[,2], z = tsne$Y[,3], Target = train_tsne_labels_binary)
plot_ly(tsne_plot_binary, x =~x, y =~y, z =~z, color = ~Target) %>%
add_markers(marker = list(
size = 3,
opacity = 0.5)) %>%
layout(scene = list(xaxis = list(title = 'D1'),
yaxis = list(title = 'D2'),
zaxis = list(title = 'D3')),
title = 't-SNE plot (binary target)')## Warning in RColorBrewer::brewer.pal(N, "Set2"): minimal value for n is 3, returning requested palette with 3 different levels
## Warning in RColorBrewer::brewer.pal(N, "Set2"): minimal value for n is 3, returning requested palette with 3 different levels# Select features
important_features = c(
"escolari",
"dependency",
"age",
"qmobilephone",
"PC1",
"PC2"
)
important = full[1:nrow(train),] %>% cbind(train$Target) %>%
rename(Target = `train$Target`) %>%
select(one_of(important_features), Target) %>%
mutate(
dependency = as.numeric(ifelse(dependency == 'no', 0,
ifelse(dependency == 'yes', mean(dependency, na.rm = TRUE), dependency)))
)## Warning in mean.default(dependency, na.rm = TRUE): argument is not numeric or
## logical: returning NAfeaturePlot(x=important[,1:6], y=as.factor(important[,7]),
plot = "box",
scales = list(y = list(relation="free"),
x = list(rot = 90)),
layout = c(3, 2),
auto.key = list(columns = 3))
featurePlot(x=important[,1:6], y=as.factor(important[,7]),
plot = "density",
scales = list(x = list(relation="free"),
y = list(relation="free")),
adjust = 1.5,
pch = "|",
layout = c(3, 2),
auto.key = list(columns = 3))
# Add split label
train_test_data = full %>% select(one_of(important_features)) %>%
mutate(
dependency = as.numeric(ifelse(dependency == 'no', 0,
ifelse(dependency == 'yes', mean(dependency, na.rm = TRUE), dependency)))
)## Warning in mean.default(dependency, na.rm = TRUE): argument is not numeric or
## logical: returning NAtrain_test_data[1:nrow(train),"Split"] = "Train"
train_test_data[(nrow(train)+1):nrow(train_test_data),"Split"] = "Test"
# Have a look at the different distributions
featurePlot(x=train_test_data[,important_features], y=as.factor(train_test_data[,"Split"]),
plot = "density",
scales = list(x = list(relation="free"),
y = list(relation="free")),
adjust = 1.5,
pch = "|",
layout = c(3, 2),
auto.key = list(columns = 3))## Warning in draw.key(simpleKey(...), draw = FALSE): not enough rows for columns
Note that the echo = FALSE parameter was added to the code chunk to prevent printing of the R code that generated the plot.